A rule-based chatbot follows paths you draw. An AI chatbot writes its own answers from your content. An AI agent goes one step further: it acts, by calling your other systems through an API (a connection between two pieces of software) to look up an order, book a slot or update a record. Most products sold today mix two or three of these, so the useful question is which parts you need and which plan unlocks them.
The short version
- Rule-based bot: buttons, keywords and branches you design. Predictable, but it only knows what you scripted.
- AI chatbot: a language model answers in its own words from your help content. It covers more questions but can be wrong.
- AI agent: an AI chatbot that can also take actions in other systems. Useful, and riskier, because a wrong step changes real data.
- Of the nine builders we review, we classify Intercom Fin, Tidio Lyro and Chatfuel as AI agents, and Botpress and Voiceflow as agent platforms.
- The label on the website matters less than two documented facts: can the bot call your systems, and who takes over when it cannot.
Three kinds of bot
| Kind | What it does | What can go wrong |
|---|---|---|
| Rule-based chatbot | Shows messages and buttons in an order you designed. Branches on answers, keywords or conditions. | Anything off-script gets a dead end or a fallback message. Every new question means editing the flow. |
| AI chatbot | Reads the visitor's question and writes an answer from a knowledge base, such as help articles, PDFs or web pages. | It can give a fluent answer that is wrong or out of date. Quality depends on your content. |
| AI agent | Answers like an AI chatbot and also runs actions: calls an API, fills a form, books, refunds or routes. | A wrong decision is no longer just a wrong sentence. It can change an order or a calendar. |
The line between the last two is actions. An AI chatbot tells a customer how to change a delivery address. An AI agent changes it. That needs a connection to your order system, rules about what the agent may do on its own, and a way to hand the chat to a person when it should stop.
Where each product in our database sits
Our product database gives each builder one type, based on what its documentation says it is built for. The table below adds what the documentation describes for flows, AI answers and actions. "Actions" means a documented way for the bot to call your own systems.
| Product | Our type | Scripted flows | AI answers | Actions |
|---|---|---|---|---|
| Intercom Fin | AI agent | Workflows builder, Advanced plan and up | Yes, from help articles, PDFs, URLs, snippets and past conversations | Procedures, data connectors and custom MCP connectors |
| Tidio Lyro | AI agent | Yes, Tidio Flows | Yes, Lyro | Lyro Actions, with a per-plan count (1 on Free, 10 on Plus) |
| Chatfuel | AI agent | No flow canvas for new accounts | Yes, from a structured knowledge base | Bookings, payment links and a JSON API plugin |
| Botpress | Agent platform | Yes, Botpress Studio | Yes, knowledge bases and Autonomous Nodes | JavaScript, an SDK, REST APIs and webhooks |
| Voiceflow | Agent platform | Yes, visual builder | Yes, with a choice of model | Code extensions documented |
| HubSpot Chatbot Builder | Bundled with a CRM | Yes, chatflows; branching needs Professional | Only through the separate Breeze Customer Agent | Agent actions set up as API calls |
| Landbot | Flow builder | Yes, the core product | AI Agent block on paid plans | Webhook and integration blocks after the AI step; no native tool-calling described |
| Manychat | Marketing bot | Yes, the core product | Manychat AI, from Pro | External Request step and an API from Pro |
| Typebot | Open source | Yes, rule-based by default | Optional blocks that use your own AI provider key | HTTP Request and Script blocks |
Two patterns stand out. First, the "AI agent" products still ship scripted tools, because some jobs need a fixed path. Second, the flow builders have added AI as one step inside a flow. Landbot's AI Agent block, for example, collects data and then exits back into the rule-based flow, where the webhooks run.
An agent is only as safe as its permissions
When a bot can act, the settings around it matter more than the model. Look for three things in the documentation. What the agent may do without review: Chatfuel's documentation describes four autonomy levels for bookings and says new bots default to full autonomy. How it hands over: Intercom Fin escalates when unsure or when asked, and HubSpot's agent hands off when it cannot answer or when the visitor asks for a person. Where the person answers: Voiceflow and Typebot have no built-in inbox, so handoff goes to another tool.
Actions also change the bill. Agents are often priced per outcome, not per message. Fin charges US$0.99 per outcome, and HubSpot's agent charges US$0.50 per resolved conversation. See chatbot pricing models for how these meters compare.
Which one do you need?
- A rule-based bot if the job is a short, fixed path: qualify a lead, book a demo, collect an address, answer a comment with a link.
- An AI chatbot if customers ask many different questions that your help content already answers.
- An AI agent if the most common questions need live data or a change in another system, and you can give it narrow, tested permissions.
- Not a bot at all if mistakes are costly and volume is low. Read when not to use a chatbot.
For the trade-off between scripted and generated answers in more depth, read rule-based vs AI chatbots. To shortlist products by job and channel, use the chatbot finder.
Frequently asked questions
Is an AI agent the same as a chatbot?
An AI agent is a kind of chatbot that can also take actions, such as calling an API to check an order or book an appointment. A plain AI chatbot only answers. A rule-based chatbot follows paths someone designed and does not write its own answers.
Which chatbot builders are AI agents?
In our database, Intercom Fin, Tidio Lyro and Chatfuel are classified as AI agents, and Botpress and Voiceflow as agent platforms for building your own. HubSpot adds an AI agent, the Breeze Customer Agent, next to its rule-based chatflows. The classification follows each vendor's documentation, not a test of the products.
Do AI agents replace rule-based flows?
Not in the products we review. Intercom Fin ships the Workflows builder, Tidio ships Flows, and Landbot and Manychat put AI inside rule-based flows. Scripted paths are still the tool for jobs that must run the same way every time.
What makes an AI agent riskier than an AI chatbot?
An AI agent can change real data, so a wrong decision can move a booking or an order, not just produce a wrong sentence. Check what the agent may do without review and how it hands over to a person. Chatfuel's documentation, for example, says new bots book appointments at full autonomy by default.
Keep reading
- Rule-based or AI chatbot: which should you build?guide
- How is chatbot software priced?guide
- What counts as a conversation in chatbot pricing?guide
- How do you write a knowledge base for a chatbot?guide
- Best chatbot buildersbest list
Researched and drafted with AI assistance from the sources listed on this page. We have not built or tested bots on these platforms yet. Method: How we review